This annotated data set contains ground truth labels of 
face tracks of six different movies. The tracks are labeled 
with gender (female/male) and age (youth/not youth). The 
movie "Love, Actually" contains additional labels that we do 
not yet use in our testing. 

Each movie directory contains subdirectories 
corresponding to the DVD VOB from which the detections
were taken.  There are three image files for each VOB.
These image files are named by their track number and
frame number and are provided for alignment purposes.

Modify the first two lines of the demonstration script 
"demo.m" to choose which movie and vob to show.  The script
will load the three image files, draw the facial feature 
points and bounding box of the face, and then show the 
facial feature descriptor patches in a separate window.

To reproduce our dataset, obtain the appropriate VOBs.
Look at the frame numbers in the named image files to ensure 
alignment. Load the training and testing splits as 
demonstrated in the second half of the demo.m file.  Extract 
the descriptor on the frame using the given facial feature 
points and the "extdesc" call in the demo.m file as an example.

More details of this annotated data set may be found at
http://www.di.ens.fr/willow/research/attribute_learning/data.html

